The Danger of Designing for a Future of Work We Think We Know

AI is changing what work looks like. The greatest opportunity lies in building organizations where people have the capacity and authority to keep discovering what’s possible, so they can help create what comes next.

We’re unbelievably early in AI, and yet we’re already designing for a future of work we think we know.

We’re predicting which jobs will disappear, which skills will become obsolete, and how many people organizations will need. Then we’re restructuring teams, investing in technology, and building operating models around those predictions. Maybe some of them will prove accurate, but others definitely won’t.

Leaders have always made decisions with imperfect information. The danger comes when we grow so confident in our predictions that we design organizations unable or slow to adapt when those predictions prove wrong.

Consider how early we really are. A chart from PNC Research making the rounds on LinkedIn shows the share of US households paying for an AI subscription climbing from close to zero in early 2023 to 2.2% in April 2026. That number measures paid subscriptions, so it leaves out the millions of people using free tools or AI their employer provides. Even so, the curve is steep and the base is still small.

The deeper uncertainty concerns what happens as more people discover what they can do with AI. That learning is happening while the tools keep getting easier to use and more powerful, and that combination is where things get interesting. People will come up with ideas and new ways of working that no strategy deck anticipated, and those ideas will ripple through companies, communities, and society.

We’ve watched this pattern unfold before. In 2005, social tools mostly belonged to early adopters experimenting at the edges. Once they became easy to use and reached mass adoption, everyday people reshaped how news spreads, how brands talk with customers, how communities organize, and how people find jobs. Most of those shifts came from millions of people figuring out what the tools made possible, far more than from the companies that built the platforms.

AI is following a similar path, but even faster. The most important changes ahead will likely come from people using AI in ways none of us have imagined yet. That makes the way we design our workforce today one of the most important leadership decisions of the AI era, because the organizations that thrive will be the ones ready to learn from those discoveries.

We May Be Asking The Wrong Questions

Instead of asking how to prepare employees for the future of work, what if we asked how to build a workforce capable of continuously shaping it? Instead of trying to predict every capability we’ll need five years from now, what if we built organizations where learning, experimentation, creativity, and innovation are part of everyday work? Instead of designing for a fixed destination, what if we designed for the ability to keep moving?

That shift also means lifting our sights from the financial bottom line toward the horizon. Many of the leaders I talk with see the biggest payoff from AI in productivity and efficiency, and I understand why. They’ve been trained to improve margins and deliver measurable returns, and those responsibilities are what keep the company alive. My worry is that we’re so focused on what AI can save us today that we’re overlooking what it could make possible tomorrow.

What new products and services could we offer? What would our people create if they had more time to think, collaborate, and challenge the way things have always been done? Efficiency creates capacity we can reinvest, and human curiosity, judgment, and creativity decide where that reinvestment goes. The future of work is something we have the opportunity to create.

Three Predictions Worth Questioning

Each of these predictions sounds reasonable, and each is already shaping hiring plans, budgets, and org charts. That’s exactly what makes them dangerous. Once a prediction becomes a design assumption, it gets built into roles, structures, and incentives that are hard to unwind when the evidence changes.

1. “We Already Know Which Jobs Will Disappear”

This assumption leads organizations to cut or freeze roles based on which tasks look automatable, often before anyone has studied how the work actually gets done. The risk runs in both directions. Leaders can lose people whose judgment and relationships turn out to be essential, and the people who remain learn that their job is to defend their current role.

Anthropic’s Economic Index, which analyzed millions of anonymized Claude conversations, found that in early 2025 about 57% of AI use augmented people’s work while about 43% automated tasks. Only about 4% of occupations used AI for three-quarters or more of their tasks, while about 36% used it for at least a quarter of them.

Think about a marketing manager who uses AI to draft campaign copy in minutes. The drafting gets faster, but the work of understanding customers, negotiating priorities with sales, and deciding which message fits the brand still sits with her. A job is a bundle of tasks, relationships, and decisions, and AI is changing those pieces at different speeds.

I recommend starting with the work itself. Map how specific workflows are changing, involve the people who do that work, and redesign roles around what you learn. That approach keeps institutional knowledge in the room and gives employees a stake in what comes next.

2. “Time Saved Equals Value Created”

This assumption is dangerous because it feels true. When usage climbs and people report saving time, leaders tend to bank those hours in budgets and headcount plans before anyone checks what the time actually produced.

The evidence on time savings is also more mixed than the headlines suggest. In METR’s 2025 trial, 16 experienced open-source developers took 19% longer to finish tasks with AI, even though they believed it had sped them up. That’s one setting with a small group, and it still shows how easily perceived speed and actual results can drift apart.

My own experience points in a related direction. I use AI for writing, for developing assessments and interactive tools, and for building upcoming training. Each project still takes me close to the same amount of time it always did. The difference shows up in the result, because I spend far less time organizing my thoughts and researching and reinvest that time in improving the work. If you measured only my hours, you’d conclude AI hadn’t helped me at all.

I recommend measuring outcomes alongside time. Look at quality, error rates, cycle time, and customer results to see where AI is creating value. When it does free up hours, decide deliberately what those hours are for, which brings us to the next question.

3. “What AI Can’t Do Today Defines What Humans Will Do Tomorrow”

Many workforce strategies define human value as whatever technology can’t handle yet. That puts people on shrinking ground, because every jump in capability redraws the map and turns yesterday’s safe role into today’s restructuring conversation.

The boundary between human and machine work is moving, and it will keep moving.

I recommend building around current capabilities while regularly revisiting where that boundary sits. More importantly, invest in what helps people stay valuable as it shifts, including curiosity, judgment, collaboration, and the ability to learn and unlearn. Those capabilities need organizational support to grow.

If AI Saved Your Team Five Hours Last Week, Where Did Those Hours Go?

Picture a familiar rollout. Employees get access to tools, training launches, and usage climbs. Tasks that once took hours now take minutes, and leadership celebrates the adoption dashboard.

There may be good reasons to celebrate. I’d still ask one question. If AI saved your team five hours last week, where did those hours go?

Boston Consulting Group’s June 2026 AI at Work research surveyed nearly 12,000 employees, managers, and leaders. Among frontline employees who regularly used AI, 42% reported saving at least eight hours per week. Yet 66% received limited or no guidance on what to do with that recovered time, and more than half said they weren’t reinvesting it in more strategic work.

Research by economists Anders Humlum and Emilie Vestergaard, published as a National Bureau of Economic Research working paper, found that workers put more than 80% of their AI-saved time into other work tasks. The time didn’t vanish. People simply did other things with it, which raises the question that matters most. Were those other things moving the organization forward?

Recovered time only becomes innovation or revenue when someone decides what it’s for. Sometimes the best choice is better service, fewer errors, or a more sustainable workload. Sometimes it’s employee development, and sometimes it’s room to explore ideas that could change the business. AI-created capacity gives us choices, and leadership needs to guide what we do with them.

Creativity And Innovation Need Somewhere To Live

One of the most valuable choices leaders can make is to reinvest that capacity in creativity and innovation. That’s harder than it sounds, because we often treat creativity and innovation as personality traits, as if some people have them and others handle execution. I believe organizations play a much larger role in enabling or limiting what people contribute. Creativity takes root when the conditions support it. People need time to explore, access to information, collaborators with different perspectives, a shared understanding of the problems worth solving, and enough authority to act on what they discover.

Harvard Business Publishing’s 2025 research with Degreed found that nearly three in four employees engage in self-directed AI learning at least monthly, yet only 12% say gen AI learning is a core strategic initiative at their organization. People are already discovering new ways to work. The open question is whether our organizations are learning from those discoveries or leaving them in isolated pockets.

Picture an employee who finds that AI can eliminate several hours of repetitive reporting. In one organization, she’s handed more reports. In another, she’s encouraged to ask why the reports exist, who uses them, and whether there’s a better way to deliver the information. The first organization becomes more efficient. The second has a chance to become more effective, and it may discover that it needs better access to information or sharper decision rights far more than it needs more reports.

We can’t ask people to imagine new possibilities while rewarding only compliance with the existing system.

Designing An Adaptable Workforce

An adaptable workforce comes from intentional organizational design. Hiring people who describe themselves as flexible helps, but the organization still has to make adaptability possible. Here are the five capabilities I’d focus on.

1. Set Direction Without Demanding Certainty

People need to understand where the organization is heading, why its priorities matter, and what outcomes it’s working toward. Leaders can share what they know, acknowledge what they don’t, and explain how they’ll revisit decisions as circumstances change. That builds alignment while leaving room to learn.

2. Make Participation Part Of How Work Gets Done

The people closest to the work often see opportunities long before they surface in a planning meeting. Give them regular ways to flag problems, suggest improvements, and weigh in on AI decisions before they’re finalized. Then act on what you hear, because participation without influence quickly becomes an exercise people learn to ignore.

3. Protect Capacity For Experimentation

If every efficiency gain gets refilled with more work, innovation has nowhere to happen. Set aside some of the time AI frees up and tie experiments to real business challenges. This is also where individual skills grow, including learning to speak to models, giving them context and direction, and evaluating their work the way you’d review a talented new colleague’s first draft.

4. Redesign Decision-Making And Knowledge Sharing

Innovation often stalls between organizational boundaries. One team discovers something valuable, and nobody else hears about it, or a promising idea needs so many approvals that the moment passes. Adaptable organizations define decision rights, look at how models and agents can support entire workflows and the handoffs between them, and organize their data so people and AI can reach reliable information.

5. Measure What Becomes Possible

Adoption and efficiency metrics matter, and they tell only part of the story. Track what teams can solve today that they couldn’t before, which customer experiences have improved, and how quickly the organization recognizes when something isn’t working and changes course.

It’s why I believe participation is the new job security. Participation can’t guarantee any role stays the same. People who keep learning, contributing, and adapting are far better positioned to stay valuable as work evolves, and organizations that make participation possible build a much broader capacity to see change coming. When employees help shape transformation, adaptability becomes something the organization practices every day.

Make Today’s Decisions But Stay Ready To Outgrow Them

We’re just getting started. AI capabilities will keep evolving, some of our assumptions about jobs and skills will be proven wrong, and opportunities we haven’t imagined will appear.

We can’t eliminate that uncertainty, but we get to decide how we prepare for it. We can design around fixed predictions about what work will become, or we can design organizations where people have the capacity to keep learning, experimenting, and creating value as new possibilities emerge.

We have to make decisions today. We also have to stay willing to outgrow them tomorrow. Commit fully to your best current thinking, but keep it close enough to examine every time the evidence shifts.

We need to keep one eye on the bottom line and the other on the horizon. Efficiency gives us the capacity to invest in tomorrow, creativity helps us discover what tomorrow could become, and adaptability lets us keep moving as it unfolds.

So here’s the question I’d leave with every leader. Are we designing our workforce to operate in the future we predict, or to help create a future we haven’t yet imagined?

I’d love to hear from you. Which assumptions are you revisiting in your own work, and what have you learned so far?

Oh, And One More Thing…

If you’d like to keep this conversation going, I’ll be speaking at several events this month. At ElevateHer on October 16, I’m facilitating a roundtable called “AI Gave You Three Hours Back. Now What?” that explores what happens after the efficiency gains. On October 21, I’ll introduce the PATH Framework™ at Atlanta Women in AI in Alpharetta, and on October 22 I’ll join the free virtual All Day AI conference to ask who is developing the workforce as AI changes the work. You’ll find details for all of them, plus the next Reflection Series conversation on October 20, on my events page.